tl;dr: Anthropic researchers used Claude Mythos Preview to discover new cryptographic attacks: one that halves the effective key strength of HAWK (a NIST post-quantum signature candidate) and another that speeds up attacks on 7-round AES by 200-800×. Neither affects production systems—HAWK isn't deployed and the AES attack only works on a reduced-round variant—but the results demonstrate that frontier LLMs can now perform novel cryptanalysis at expert-researcher level, with each discovery costing roughly $100,000 in API usage.